[en] Although the discovery of biological epistasis via statistical methods remains a big challenge, large-scale epistasis studies have become more and more popular nowadays. Considering the power of genome-wide association interaction analysis (GWAI) is conceptually lower when comparing to GWA using the same data, the genome-wide screenings for epistasis often ends with no statistically significant findings. Therefore, sound meta-analytic methodologies for combining evidence across independent gene-gene interaction studies, taking into account characteristics of individual studies, as well as protocols for good standard practice of meta-GWAI analysis are missing so far. The existing methodologies, whereas is suboptimal, may still be the first choice in meta-GWAI. In our study, the genome-wide significant interaction between two var¬¬iants of WWC1 and TLN2 genes were first discovered in males of EADI1 of Alzheimer’s disease (AD) case/control cohort. Adopting the gene-based replication approach recently advocated by Gusareva and Van Steen (Hum Genet 2014), we assessed association between all available variants from the pair of genes to test for pair-wise interactions in three independent AD replication cohorts (GERAD1, AD cohort from Rotterdam study - RS, and ADGC). Particularly, we used regression modeling with co-dominant coding of genetic markers (the one that was found to be the most powerful to identify epistasis and gives the least false positive results) and taking into account possible confounding factors (age and the first four PCs). We further meta-analyzed p-values for the interaction effects across all pairs of markers in males and females using Fisher summary test statistic. For the most interesting epistasis signal that occur in males of EADI1 cohort and two replication cohorts (RS and ADGC), we also conducted multi-locus genotype (MLG) association analysis (quantifies effects sizes of each of eight multilocus genotypes derived from the SNPs-pair versus the reference category - homozygous for the major alleles) and then met-analyzed results using the random effect or fixed effect models depending on presence of heterogeneity between studies. Adopting this approach we statistically replicated the interaction between WWC1 and TLN2 genes. Although, the biological validation of this finding is still ahead, we can already demonstrate effectiveness of the adopted meta-analysis pipeline to identify novel genetic factors potentially predisposing to AD.
Disciplines :
Genetics & genetic processes
Author, co-author :
Gusareva, Elena ; Université de Liège > Dép. d'électric., électron. et informat. (Inst.Montefiore) > Bioinformatique